Microbial, Behavioral, and Clinical Correlates of the Gut-Brain Axis in Anorexia Nervosa: Insights from the ANIGMA Cohort

Anorexia nervosa (AN) is increasingly conceptualized as a metabo-psychiatric disorder, characterized by severe metabolic disruption and psychological symptoms. However, the extent to which gut microbial composition reflects metabolic, psychological, and clinical variation in AN remains incompletely understood. In this study, we investigated the associations between genus-level gut microbiome composition and metabolic, psychological, and clinical variables in participants from the Anorexia Nervosa: Investigation of the Gut Microbiome and Anxiety (ANIGMA) cohort. Participants were young adults (mean age AN: 26.3 years; non-ED controls: 24.9 years). Stool microbiome data were analyzed from individuals with AN at admission (AN-AD; n = 118) and discharge (AN-DIS; n = 98), as well as non-eating disorder controls (non-ED; n = 132). Following preprocessing, metadata alignment, and analytic exclusions, 189 samples were included in the present analysis. Non-ED controls had no history of BMI outside the range 18.5-24.9 kg/m² or any ED. Microbial relative abundances were log-transformed and evaluated using genus-wise linear models adjusted for age, with predictors including body mass index (BMI), body weight, perceived stress (PSS), and anxiety measures (STAI_TOTAL, STAI-Y1, STAI-Y2). Multiple testing correction was applied using the Benjamini-Hochberg false discovery rate (FDR). Differential abundance across timepoints was additionally assessed using ANCOMBC2.             BMI was the only variable associated with microbial composition after multiple testing correction, with three genera meeting the FDR threshold at q < 0.05. Specifically, Roseburia and [Eubacterium] siraeum group were negatively associated with BMI, whereas Gordonibacter was positively associated. In contrast, body weight, perceived stress, and anxiety measures demonstrated multiple nominal associations (p < 0.05), including consistent signals involving Gordonibacter and [Eubacterium] siraeum group, but no taxa remained significant after FDR correction. Across all pairwise comparisons of groups (Non-ED control vs AN-AD, Non-ED control vs AN-DIS, AN-AD vs AN-DIS), no genera were differentially abundant following correction using ANCOMBC2, indicating an absence of robust genus-level microbiome shifts across disease state or short-term treatment.             These findings demonstrate that BMI was the strongest and only robust measure of gut microbiome variation in this cohort, as psychological measures and disease status exhibit only exploratory or non-significant associations. The lack of consistent treatment-dependent differences raises the possibility that genus-level microbial composition may not undergo large transformations during renourishment, or that such effects are subtle relative to inter-individual variability. Future studies should aim to integrate longitudinal within-individual analyses, higher-resolution taxonomic and functional profiling, and multi-omics approaches. These data will be critical for conceptualizing and clarifying the role of the gut microbiome in both the development and recovery of AN.